Old repair technician chiming in, he said "make sure no one else have attempted to repair it". Those words right there - spot on! The worst repair cases I ever had is from incompetent repair attempts and you will NOT believe the stuff I've seen. I've seen attempts where people have literally FORCED their soldering iron right through a 7 layer PCB board (that's unrepairable right there) and the unit is destroyed forever. It almost always pays to purchase units that no one has attempted to repair before, but sadly - so many dishonest people just sell their failed repair attempts.

"I just bought 7 dead gpus for _only_ $1700" Sir you and i are still on very different playing fields

"It's working but we've got no idea what we did..." Years ago, AT&T had an official resolution code for their techs, when we'd call in a line problem that just started working while they were troubleshooting: FWT = Foxtrot Whiskey Tango = Fixed While Testing :D

$400 for a damaged 1080ti is WILD 😅 I know the videos 8 months old but sheesh

The fact that a broken rx 580 in these days is more expensive than the one i bought used 3 years ago is nuts

I like how Linus warned nearly every intel tech upgrade recipient that their PSU was a little underspecced, yet origin here is casually throwing in 850W for a 3080Ti and 12900K

My job involves testing very sensitive electronics. When using a DMM in auto-range mode (for resistance or continuity mode), the DMM can actually put out quite a lot of voltage that can easily damage parts not rated for that voltage. When the DMM is set to a low resistance range, the output voltage could be sometimes as high as 9V. With modern chips using core voltages as low as 1.2V, 9V is way too high and can easily kill the chip. When the DMM is set to a high resistance range, the voltage will be much lower and will be at a safer level. When set to auto-range, which is the default setting, you really don't know what you'll be getting as the output voltage. If you don't know how much open circuit voltage your DMM has and you need to check for a short, it's recommended to set the DMM to the highest resistance range and use resistance mode. When in a high resistance range (like 1M or 3M), a short will read 0 and a non-short will read at the low end of that range like 1M. The beeping continuity mode on a DMM will likely put out max voltage so its not recommended to use. On many DMMs, the diode mode and beeping continuity mode are the same so only use it if you are checking for whether a diode is in series has the right voltage drop or if you are checking continuity on some passive wire. One thing you can do is check the open circuit voltage of your DMM to see if it's safe enough for general use without having to set the range first. We use really old Fluke 77 DMMs because they haven't low voltage.

What’s funny about the sticker thing is they did a whole video about how to take those kind of stickers off and put them back on without the manufacturers being able to notice and deny a warranty

I worked at a computer repair store for awhile, and it seems that the "dead" electronics people brought in were almost always split into two categories: Not really dead and power-cycling brought it back, and so dead that we didn't have the time, budget, or equipment to fix it. I didn't really get any satisfaction out of either category.

Would love to see more "Can We Fix...?"-type videos. With component prices all over the map, logistics shutting down every other month, and other... uh... geopolitical issues, this could really save someone's bacon. Odds aren't great -- it is eBay, and the parts are explicitly listed as broken. Still, you'll miss 100% of the shots you don't take. Also, even a bad board might be worth something to the would-be repairer, for practicing more advanced soldering techniques (like hot air reflow). It's not like you'll break it _more,_ right? 👍️

One reason that taking it apart and reassembling fixes stuff is that oxidation on connectors is scraped off when you unplug and plug again. For extra cleaning, remove and insert several times. You can also use a pencil eraser on edge connector fingers to remove oxidation.

About the oven trick: It's definitely worth a try. I've had 2x 780 ti for years where one of them stopped working, I fixed it with the oven method which made it last 3 months, fixed again and it worked 2 months more then it wouldnt work again. My second card stopped working later and after I ran it in the oven it has been working for 8 months right now, hope it will last more :)

RE: Jono's issue. Had the same problem with a monitor falling off. Replacing the HDMI cables addressed it.

Love this calm competent approach. Doing some "can we fix" themed episodes would be great. Aware hardware gets harder and harder to fix but what is worth giving a go instead of default to throw out out and replace.

Turn off the Multimeter when putting it away after measuring. Not only will it preserve the battery but you won't use it in the wrong mode by accident (and possibly damage it)!

This is one of the cooler episodes from this channel. Would seriously love to see some more of these.

PSA : When trying to find shorts on either Vcore or Vmem, never ever use continuity tests ! You'll get beeps even on a 100% working card because the loads are so small (a couple ohms maximum). You should use normal resistance measurement and make sure that you get a "true" ~0 ohm reading

It’s still mind-boggling to me what has happened to the GPU market. Their broken RX580 cost them $150, meanwhile I spent $120 for my (perfectly fine) one a couple years ago

22:47 my guess would be there was a slow leak somewhere else in the water loop that dripped onto the GPU over time, probably in the CPU block

This was really interesting. I wonder how far you can go tho, I mean that card with the melted IC on it looked like it was screwed for sure but it would be interesting to see just how much you can salvage if you really go all in for it

2:38 The hardest failure to diagnose is 'intermittent' failure! If you fail to pinpoint the problem and decide to 'look around' and test components, you may end up bricking the whole thing. Most intermittent failures are faulty boards: cracked or semi fried... I always avoid touching those

There's too many shady sellers that KNOW something is beyond repair but sell it for a ridiculous price hoping some person will want to take the gamble. Even for parts, these cards don't come anywhere CLOSE to what they ask for, it's ridiculous. Not to mention so many broken items have already had work done ruining them beyond repair.

More of these videos please! They are incredibly entertaining while still being educational:)

I've seen the same issues with video cards cutting out when over loading the power supply. Either the power supply itself is not strong enough for everything in the system. The other big issue is over drawing one rail. Most of the newer graphics cards need to have power supplied from two different rails/outputs from the power supply.

7:34 you gotta be careful of this... don't clean a card right away. Take note of all the spots with corrosion and look at them in depth before cleaning. they can be a good hint as to where the issue may be.

5:20 one thing to note is opening up netflix or any other service using HDCP will temporarly black out your screen as the protocol change, and some monitors might just have trouble with HDCP too

I’ve done the oven GPU fix on a number of occasions, especially on 2010 IMac and I can vouch it works. I don’t think it reflows this old. Are your correct on that one. But but it does definitely affect the soldier in someway. It’s the only possible thing they can be. I know dry shoulder joints or something people talk about so possibly something to do with that

I am a total rando with zero tech background but I love fixing old broken things around the house that would just get thrown away, I enjoy doing that and after I am done I take it to the nearby shop that buys second hand stuff and sell it hella cheap and get some cash for my troubles. I find videos like this very fun to watch for the same reason. I have learnt a lot over the years watching videos and just meddling with broken stuff.

The first GPU seem to me like a Power Supply issue, by experience having the same issue before and knowing someone who had the same issue too, both of our PSU ended up dying.

Gotta love this 20 minute demonstration of how fixing broken tech is basically a D&D "skill check" where you roll a d20 and you can succeed or fail purely based on chance because at some point logic no longer seems to factor into the problem solving equation lol

"I'm gonna zip tie this fan to it, hell yeah." This killed me.

For Jono's GPU, I've had that issue and it was just a bad HDMI cable, or one that was not rated for the bandwith. Swapping it resolved in my case for a GPU that would randomly loose sync/drop video or loose sync in an observable pattern.

can we briefly talk about how crazy expensive these broken cards are? it should not be more than 10% of the new card price

A card that suddenly starts working just from taking it apart and cleaning it likely had a bit of metal dust or filings or corrosion sucked into it that shorted something, dust or other crud causing a support IC to overheat, or the thermal paste was so bad that the GPU die overheated nearly instantly and go into self-protection shutdown to avoid melting. Cards with shorted power rails are usually quite easy to fix; scrub the card down to knock lose any corrosion or conductive dust that could be shorting it, and replace a blown MOSFET.

I was given a 980 TI that seems to run on 1 x 6 pin connector, but the other one does not seem to allow it to run. I've been trying to learn more about this stuff to trace it myself, but I do appreciate videos like this to help me learn what to look for.

I don't know a lot about GPUs, but when you mentioned it being time based on PCI-E, sounds like it's impedance controlled. That means any change to length, type of connection, or even the soldermask on top of it will affect the response times. If it's really like that, then it's most likely a goner unless you have some sort of copper welder for traces (which are half a million or more for the super special ones we'd use for my line of work).

A short on the 12V input is actually one of the most fixable faults, most of the time a dead mosfet and a good chance the gpu itself survived.

My theory for the card with the excessive water damage: The CPU in the original system was probably water cooled aswell and may have leaked onto the GPU.

6:32 I once had a huge short in the ATX 12V connector on the top of my mobo, and it didn't blow my Seasonic PSU. It was a zero-Ohm short but the thing was intelligent enough not to burn itself.

You guys should get a handheld LCR meter. It will allow you to measure all the caps while in circuit.

the thermal pads evga used for the "fix" for the FTW cards of 1070 and 1080 really seemed to leak grease stuff from the pads EVERYWHERE on the cards. I dont think something was spilled in that card. I recently replaced the thermal pads in my 1070 ftw and had the sticky stuff on it as well. It really let dust stick to it.

I’d love to see this as a series, vocational engineers make things just work

I fixed my gtx 1660 which had a failed vram chip (confirmed using MODS). I bought the gddr6 module from aliexpress, which was the only place I could find it for sale. Used hot air to remove the old one, and to solder on the new one. Card is still working great seven months later.

2:20 it looks like it's problem might be that it punched him in the face tbh...

The PCI-e differential pair length matching requirement you mentioned isn't as tight as you think. For Gen3 PEX you've got a 20-80 rise/fall time margin of 19ps (i.e. on a transition the the signal must go from a 20% voltage level to 80% voltage level, or vice versa, within 19ps), which means a timing skew budget of +/-9.5ps. On FR4 dielectric that means you're looking at about a +/-1.7mm length skew budget before you go out of spec. Keep in mind, though, that "out of spec" doesn't mean "not working". If you blow that budget by an additional 50% you might still get a perfectly functional card. Bodge wires should be fine here if you're careful about length and keep it close to the surface, so it's 100% worth an attempted fix. One thing to consider is that timing skew isn't just about length. The amount of timing skew in the transmission line is proportional to the impedance of that line, which is proportional to the distance AND the dielectric constant. When you've got an impedance-controlled PCB, the dielectric constant can be considered (unsurprisingly) a constant, which means (with careful design) the impedance of your traces, and therefore propagation delay through them, is largely just a product of length. I'm glossing over some details like fibre weave effects, copper roughness, etc., but the main thing here is that the trace is a fixed distance away from the ground plane, with a known material (the FR4 fiberglass) between the trace and the plane. Why am I bringing all of this up? Because timing skew arises from a mismatch in impedance, and impedance is affected by the dielectric constant, and the dielectric constant is affected by the trace's distance from the ground plane and the materials that are in the way. (Side note: one thing that a lot of people get wrong here is that the distance between the two traces of a differential pair DOES NOT matter anywhere near as much as you think it might on a PCB. In a twisted wire pair the cross-coupling of the fields between two sides of a differential pair is very important, but on a PCB those two traces aren't twisted together, so the majority of the field energy is between the traces and ground plane. The two traces act much more like single-ended signals, with a small amount of cross-coupling of the fields, than proper differential signals in a twisted wire pair. The signalling also isn't truly differential in most cases with protocols like PEX, LVDS, DDR, etc. - TL;DR the lengths of the two traces must be pretty closely matched, but a consistent distance between the two traces in the differential pair is far less critical.) When you lift a high frequency signal away from the PCB surface, through a bodge wire, you alter the impedance of that trace in a way that isn't just proportional to distance. As such the length skew isn't the only factor to consider. Lifting the trace into a bodge wire also creates an impedance _discontinuity_ (i.e. an abrupt change in impedance), which causes signal reflections that can reduce the quality of the signal and close the eye diagram. These may be small effects or large effects depending on a whole bunch of factors. It _may_ be a problem. It _may_ not be a problem. For a short distance like this, my intuition is that the discontinuity might be fine on Gen3 - if it were Gen4 PEX I'd be much more concerned. One of the main sources of impedance mismatch will come from the repaired trace no longer being closely referenced to the underlying reference (ground) plane, since it's now a floating wire. So, if you want to get real fancy, you can scratch a bit of the soldermask off next to the locations where a bodge wire is going to/from, then solder a second piece of enamelled wire to those exposed ground points and wrap it around the bodge wire (or rather wrap them both around each other like a helix), forming a twisted wire pair where one side carries the signal and the other is referenced to ground. This ensures that the return currents for the lifted trace remain in the ground plane, that you get excellent common mode noise rejection, and that the fields stay tightly coupled rather than spreading out and causing radiated EMI problems and cross-coupling. It'll also serve to slightly curb the problems caused by the impedance discontinuity. This might not be required but it's a good option to have in the toolkit. I know I braindumped a bunch here and it's a bit of a daunting topic, so if any of you at LTT want to chat more about this stuff you can drop me a DM on Twitter (gsuberland) or IRC (same handle, Libera network) and I'd be more than happy to talk through this stuff and clarify anything that's unclear :)

I had an old dead 770 about 5 years ago and saw the video Linus made about putting it in the oven, I figured its a card I don't care about and had nothing to lose at this point so I gave it a shot. To my complete surprise the card worked fine and still works to this day.

Returning to this channel again. Hello. This host is super cool and entertaining to watch. Love the variety of personality on LTT. You're awesome.

That fan spinning tip. On point. ❤

Wow I didn’t know that spinning the fans that fast could be bad. I guess I won’t do that anymore. Thanks for the tech tip!

"A lot of the time, just, Disassembling and reassembling stuff that's dead, is the easiest way to fix it. You won't know why you fixed it, but it just does" I cannot POSSIBLY tell you how many times this has happened and worked for me. After cleaning, and rebooting, I just take stuff apart and put it back together, and then it magically just works. No answers, just solutions. Also, the number of horror stories of trying to work with something someone tried to repair in the past, only to find it completely destroyed beyond repair: Insurmountable. Thank you for this video. I absolutely loved it as a Professional Technician and Hobbyist Engineer.

The card with the fried Mosfett would still be a relatively easy fix. With the Mosfett removed check the capacitors around it for a direct short. Any that short replace and then new Mosfett should be fixed.

I feel like sometimes the sticky shit on the pads gets all gummy and sort of runny after time. An nvme drive died on me once because the sticky foam pad that held the cover on got all gross over time and spread causing a short.

I'd be willing to bet the one with severe water damage was in a system that had a custom loop on the CPU and THAT leaked at some point into the GPU, killing the card. All that white crusty shit is likely the algicide/minerals in the open-loop coolant.

24:32 I could have used one of those when I bricked my card with the wrong bios! It was years ago, I had a r9 270 I was having issues with so I tried updating the bios. Turns out Techpowerup doesn't always label their bios correctly, they had a r9 290x bios on the r9 270 page. Always double check your bios before you flash it

You can deal with differential traces by re-routing the whole path (where possible) from A to B for paired matches on a daughter board. Have a friend who did this several times and it works. Best part is, the daughter board can be a discard flex cable of the same value route this also matters). If this path is obstructed through the board, not a surface path, your chances for making it work go from "possible" to "mission impossible". In short terms you have to have a good microscope, measuring means, and lots of soldering skill, because even that counts.

17:50 I had an 8x card that needed to go into a board that only had a 4x slot. Just took pliers and broke off 4 of the lanes and it worked a charm

Bloody hell, we live in a time when a 4GB RX 580 sold as not working can be sold for $150 and considered a good deal if you can fix it, when just 3-4 years ago I bought a 8GB model for just £100.

I never considered that the fan spinning when cleaning could turn into a generator lol, good tip, I used to spin those suckers up with air at max rpm when cleaning because it sounded cool >_>

Just commenting on your last bit of the video (oven method). I bought a GeForce GTX 1080 Xtreme Gaming 8GB for $100 on marketplace (labeled as used- good and promised that it was working- I knew it wasn't gonna work well). Lo and behold, artifacts! My PC crashed constantly. Would restart and blue-screen on me. I cleaned it up, applied new thermal paste, new pads, and still had artifacts. I did the oven method and it worked with 0 issues. I have done this to 3 different cards and they're still gaming strong with no issues a year+ later. It may be luck, or it may be that it actually works most of the time. Idk, but so far..It's been a good cheap method to fix GPU's that have been artifacting. I've never had any issues that people bring up to debunk it or that it stops working after a short time.

Components that "just" have been under water can mostly be repaired by just cleaning them. A friend of mine had a flooding in his area and his office was halfly under water. All of the computers were full with this agressive mixture of water, mud and whatever came up from the seweage. Luckily almost all computers were turned off. I took them all apart (wearing gloves and a mask) and cleaned them with disinfectant, dish soap + distilled water, using a brush and finally some fresh distilled water again. The cases were cleaned with a garden hose and disinfectant. I didn´t open the power suplys, because that is dangerous, but some of them were just above the water. I let it all dry for a week and put them back together. They almost all worked! I just needed to put in some few spare parts.

Alex: Nice tip on the fans,(cleaning) however, you didn't mention that the fans bearing has a specific rating based on lubrication and surface. You can very easily take a fan beyond it's speed/heat rating with a compressor. Start it back up and your fan doesn't work, or squeaks? Probably killed a bearing.

I love watching videos like this. Seeing people troubleshoot GPUs and finding one dead component like SMT resistor or capacitor looks like wizardry.

I like your attitude Alex, I think we'd get along lmao, I've always said the same thing about disassembling and reassembling stuff, it's crazy how many times you can fix something and not even realize how it happened

Never had a gpu live past 1 year the day after running furmark. RIP my 2900 xt, gtx 460, and gtx 560. Definitely spent too much time blaming it on other problems and I've only done mild overclocking.

The "liquid" that you thought was spilled on it might be the thermal pads leaking oil. I have seem many cards that have those marks and they were from bad pads that were leaking all over the card.

About the Solder Stove Repair: I did it while with the infamous Nvidia Series which didn't had flux at all and bad crystalisation Problems: You heat up to 50°C then to 80° when it reached it you shut the stove off and wait till it reached about 40°C that solves the crystallisation problem and helps for the Cracks. When it works again replace the thermal paste and it is good for some time. The solder cracks and crystallisation problems comes from uneven heating and cooling. So keeping the card better cooled may keep it save for 6 month to 1 year. Without renewing thermal paste: 3-4 Weeks.

My favorite memory of a fix was when I got a rental dvd stuck in the player. Basically the door didn’t want to open when I pushed the eject button. Opened it up, didn’t unplug it, got shocked real good, and the door just glides open. Worked fine for another 8 years.

As someone who currently works in electronics assembly i can tell you a few things about the boards you should know. the video card with the ball of solder on the sideof the resistor is a normal process thing and is usually left here as it doesnt hurt anything. the "mossfet" blowing up you would be better off getting that off the board and just bridging it. leaving it on there you run the risk of the power going down the signal line. the "hail mary toaster oven bit" has real merit if you can control it well. typical leadfree solder melts at 221C leaded is 183C there are lowtemp leadfree but those are not usually on performance electronics. Optimal oven process 230c 45sec in a nitrogen atmosphere. This should be enough to Reflow the solder yes thats the correct jargon. If there are BGA parts on the board they need to protected from collapse _this is the biggest risk here. The nitrogen will prevent the oxidation due to missing flux. also leadfree solder is at risk for tin whiskars.. that is a huge conversation all by itself. aerosol pcb cleaners are your best friend here. soft brush and air to blow dry. have fun getting zapped!

Great segment! I really enjoyed learning more about the digital side of circuitry. I have a basic background in electronics. And what I like the most, was how you got right to the bones of the caus/ problem. You addressed when you should give up and why. I was very intrigued to learn caps, had to be equal distance away from the terminal. & probably best to leave that kind of fix to a machine or a professional perfectionist. I would very much like to see more segments about fixing computers, boards, & electronics in general. Perhaps you could add it to your upcoming test channe or do more of them on this channel. I could probably find a channel like that, but I usually tune out when it gets too technical & drawn out.

13:46 Miracle Max approved. Interesting video. I'm continually fascinated by how many things can be brought back to life from "dead" with some simple disassembly, cleaning, and maybe a few bucks worth of components here or there. Would happily watch more of these. :)

The transistor on that 1080 is meant to switch on/off. Not just stay on. I do competitive analysis on hw. Bet I could figure out what that transistor does and how to fix it. The chip/memory are honestly still probably fine. That transistor might be for a voltage regulator so you're probably getting a screwed up voltage I'd imagine.

Flux isn't required for soldering, it just makes it work a lot better, so reflowing works. And I've recovered a dcdc charger that had a full glass of rum and coke spilt on it then was used for another week successfully. Water damage isn't always the end

I love these repair videos. Its always good to see abandoned tech brought back to life. Y’all should do more!

$1700 for 6 older busted GPUs...that's pretty rough.

My GTX 970 is stuck at 8x as it is missing several capacitors beyond the 8x mark of the PCIe connector. Still works fine however (just uses a lot of power for a 970 - over 200W stock). Fixed a few old AGP cards with blown caps and they are all working fine...except an old Sapphire Radeon 9200 and 9200SE - overheated the DRAM while replacing caps near it due to heat flowing through ground. Not a huge loss as the SE was a terrible card and the 9200 one of the slowest 128bit versions. I've had far greater success with old motherboards - however the size of ground planes can be an issue if you don't have a good desoldering station.

The same GPU crash issue actually happened to my old pc's GTX 750 TI, but once we repurposed it as a file server for my house, there have been no issues.

To me, Jonos problem sounds like a grounding error somewhere between the monitor and the pc. (probably the monitor i'd wager) sending electricity where it shouldn't go and knocking it out for a few seconds. I have a similar problem with my USB cables at home, and anytime that my cat goes near my computer something disconnects randomly..

I’d like to see if they could buy a few gpus being sold just for parts of the same model and try to make one fully functioning one.

I liked the episode, good stuff. There will be plenty of used mining GPUs on the market after the Ethereum merge I would imagine.

In my experience, random crashing with GPUS is often caused by old thermal paste. Sounds too simple but I've been surprised too many times now. Even though the temps were fine.

This was great,Id love to see more attempting to fix dead cards etc,Especially more exotic and weird cards like Quadro,Titan,Radeon Pro etc.

2:33 Is that a black eye there on Jono?

I actually found the best fix for anything tech, during the ride from your house to some kind of repair shop things just happen to fix themselves

there must be a flyback diode or your board would fry every time the fan motor turned on or off. the Motor generates a magnetic field that will travel back on the wire and damage the transistor that turned it on or off.

Hey, i have some info that might help. If there is a short between the 6pin or 8 pin (or even the 12V or 3.3V pins of the pcie connector) it's very common that the card has a broken phase, in particular mosfets are the ones who fails the most, and you can test them probing the two sides (if i remember correctly, you have to probe between gate and source but i'm not shure). In that case, you have two options: replace the mosfet with a heat-gun or just cut off that phase by ripping off the broken mosfet. You have a chance that the chip is still alive, and if you choose to cut one phase the card will work but it will have some stability issues so you have to downlock it (it depends on how strong are the other phases). You can also try to make the remaining phases a little bit stronger by adding some capacitors or replacing the inductors with bigger ones. Anyway, hope I was helpful and I loved this video! (PS, the 1080 probably crahed because the broken mosfet doesn't make a short, but it's still not working so the card doesn't have enought power under load)

Jono's problem can easily be HDMI cable breaking... I had similar issue with my gpu so I swapped the cable to ps4 and then PS4 had the same issue so I replaced the cable completely... It also can be monitor firmware that also happened to me with brand new hdmi cable Or it can be caused by wrong setting in either monitor settings or gpu driver settings... Without his exact setup it's impossible to tell

literally just reminded me to update my driver so thank you!

20:27 Oh hell yeah. Zip tying a noctua to anything is always effective. 23:01 Oh yeah. That's a classic. Desktop sits on floor under the overhang of the desk, and a drink spills on the desk and pours directly into the top of the case and onto the graphics card.

They sat they bought these GPU's, deep down we known Linus just dropped them all one too many times.

As if Luke fixing a dead CPU was bad enough. Leave it to Alex to take on the hardware jobs no one else at LMG would.

I love that these videos just feel like we’re secondary to him and David’s conversation. Like we’re the youngest sibling in-between a conversation of the older siblings, occasionally recognized.

I had a 2011 iMac, and the GPU eventually died. The oven trick though worked a treat, and it lasted another year before giving up the ghost. I suspect if I wanted to go through the pain of pulling it all apart again, another oven session might bring it back again, but given you have to pull the entire thing apart to get the gpu out, I'm not keen.

kyle is actually great fun to have on camera, hope to see more of him in the future!

Really awesome video! As a electronics technician this was super interesting, as I usually only work with custom hardware. I would really like more of Kyle in the future :)

Updating monitor's firmware ... damn never thought you can do that! None of those cards should have been bought though, it should have been a throw away pile sent only for shipping costs or at best a few bucks for the hardware value. Great video.

I never knew about not letting your fans spin with compressed air....that's the most fun part lol

12:11 Short circuit > blown MOSFET > blown fuse (probably exist) and everything else is fine. For best case sceanrio, there is a fuse / choke for each phase, then even the card is work in some level.

One tip I can say is to take pictures of the hardware before cleaning it because if you end up taking it to a technician, it can make the job more difficult because they can use the mess as a hint of the problem Louis talks about it a couple of times in some old videos

I love these videos so much, I could watch them for hours

I got a free ps4, completely dead. What sat around with a mate, completely disassembled it, removed some plastic parts from the board, stuck it in the oven at max temp 250C for 30minutes. put it together again for a laugh and the bloody thing worked. still works to this day.